Research on Multimodality Face Antispoofing Model Based on Adversarial Attacks

<div>Choosing different numbers of image patches for white-box and black-box attacks. The abscissa represents the number of attacked image patches, and the ordinate represents the detection accuracy of the model in the face of the attacks. The first line shows the results on RGB images and Depth images, respectively; the second row shows the results on IR images and multimodality images, respectively. (a) RGB. (b) Depth. (c) IR. (d) Fusion.</div>

Security and Communication Networks

fig9

Figure 9

Figure 9: Research on Multimodality Face Antispoofing Model Based on Adversarial Attacks 